Operational problems in growing organisations rarely announce themselves.
Operational complexity grows with the organisation: handovers that fail, identical sites that diverge and procedures nobody follows.
Most leadership teams encounter operational complexity as a series of unrelated irritations. A handover that fails between two departments. A site that produces different results from an identical site. A procedure that everyone agrees with and nobody follows. A quality problem that was fixed last year and has returned in a slightly different form.
Related advisory: operations consulting and operating model consulting.
Treated individually, each of these looks like a local failure. Someone was careless. A team is under-supervised. A system needs replacing. The response is usually local too: a corrective conversation, a revised document, an additional check. And because each response is reasonable, the pattern connecting them stays invisible.
Four patterns explored in this hub, in brief.
The most common assumption about operational complexity is that it grows in proportion to the organisation.…
Every organisation of any size maintains a description of how its work is done: procedures, manuals, workflow…
Senior visibility of operations declines as organisations grow, and it declines faster than most leadership…
There is a pattern in operations so consistent that it deserves to be treated as a structural feature rather…
Open any heading to read the full discussion.
The most common assumption about operational complexity is that it grows in proportion to the organisation. Twice the people, twice the coordination. In practice the relationship is considerably less forgiving, because what has to be coordinated is not people but the connections between them.
A team of six has a manageable number of working relationships. Adding a seventh does not add one relationship; it adds six. The same arithmetic operates on functions, sites, product lines, systems and customer segments. Each new element has to be reconciled with everything already present. This is why organisations frequently report that a period of growth felt disproportionate — that adding thirty per cent more people appeared to add substantially more than thirty per cent more difficulty. The perception is not exaggeration. The load genuinely rose faster than the headcount.
Jay Galbraith’s work on organisation design, published in Interfaces, framed this as a question of information processing. His argument was that the form an organisation takes is driven by how much uncertainty its tasks contain, because uncertainty determines how much information has to be processed during the work itself. The more that has to be worked out while the work is happening, rather than settled in advance, the more the organisation has to spend on moving information between the people doing it. Growth increases uncertainty in exactly this sense: more variants, more exceptions, more parties whose knowledge is partial.
What makes this difficult to see from a senior position is that the cost is distributed. No single coordination act is expensive. A clarifying message, a short meeting, a check with another department before proceeding — each is trivial, and none appears on any report. The cost is in the aggregate, and the aggregate is borne by the people doing the work rather than by the people reviewing it. Leadership teams often discover the scale of it only indirectly, through the observation that experienced staff seem busier than their workload should warrant.
Coordination load does not distribute evenly. It concentrates at the boundaries between parts of the organisation that have different priorities, different vocabularies or different measures of success. Research by Paul Rogers and Marcia Blenko, published in Harvard Business Review in 2006 and drawing on a Bain survey of executives at more than 350 global companies, located decision-making stalls at precisely these interfaces — between functions, between the centre and business units, between global and local, and between an organisation and its outside partners. The pattern their work describes is that stalls cluster where authority meets a boundary, not where individuals are weak.
This is a useful corrective to the instinct that identifies a bottleneck with a person. Bottlenecks that sit at interfaces move when a person is replaced, because the interface remains. It is one of the reasons operational problems have a reputation for reappearing in a different location after being solved — a pattern examined further in why bottlenecks move instead of disappearing.
There is a second mechanism at work, and it is less intuitive. The structures an organisation builds to manage coordination become, in time, a source of coordination load themselves.
A weekly meeting introduced to resolve a genuine gap between two functions is, at introduction, straightforward and useful. Two years later there are eleven such meetings, each individually justified, and the aggregate consumes a significant proportion of the time of the people most needed elsewhere. An approval step introduced after a costly error prevents that error and also slows every subsequent transaction, including the great majority that carried no risk. Larry Greiner’s work on organisational growth, first published in Harvard Business Review in 1972 and republished as a Classic in 1998, made this the central observation: managers repeatedly find that what solved the problem of one period becomes the problem of a later one. The mechanism did not fail. The organisation moved past the conditions it was built for.
This matters for how operational difficulty is interpreted. A control that is producing friction is not necessarily a bad control, and the person who introduced it was not necessarily wrong. It may simply be a correct answer to a question the organisation is no longer asking. Distinguishing between the two is genuinely difficult, and organisations that cannot make the distinction tend to accumulate controls, because removing one requires an argument that nobody is well positioned to make.
In practice, organisations often find that the number of controls only ever moves in one direction. Adding a step has an identifiable sponsor and a clear rationale. Removing one has neither, and carries visible personal risk if anything subsequently goes wrong. The asymmetry is behavioural rather than deliberate, and it is one of the more reliable features of the territory.
Every organisation of any size maintains a description of how its work is done: procedures, manuals, workflow diagrams, system configurations, induction materials. Every organisation of any size also does its work somewhat differently from that description. The interesting question is not whether a gap exists — it always does — but how wide it is, in which direction it runs, and whether anyone knows.
The distinction between work as it is imagined and work as it is actually done was developed by Erik Hollnagel and colleagues, and set out with Robert Wears and Jeffrey Braithwaite in a white paper published jointly by the University of Southern Denmark, the University of Florida and Macquarie University. Their central observation is that things do not generally go well because people follow procedures exactly. They go well because people make sensible adjustments to the demands of the situation in front of them. The adaptation is not a deviation from good practice; in a great many settings it is the mechanism by which good outcomes are produced at all.
That reframing has an uncomfortable implication for organisations that treat the documented process as the authoritative account. If the written procedure describes a version of the work that has never quite existed, then improving the procedure improves the description rather than the work. And if the adjustments that make the work succeed are undocumented, they are also invisible, unprotected and non-transferable.
The gap opens through ordinary, unremarkable mechanisms. Procedures are typically written at a moment of relative stability, by people with a clear view of the intended sequence and a partial view of the exceptions. Conditions then change: volumes rise, a supplier is replaced, a system is upgraded, a category of customer arrives that the procedure never anticipated. The people performing the work adapt, because they have to. The document does not, because nobody has an obligation to update it, and because updating documentation is work that is visible only when it is absent.
Local adaptations then propagate unevenly. A useful adjustment developed in one team spreads to adjacent teams by conversation, not by publication — which means it spreads as far as the conversations reach and no further. This is how two parts of an organisation running what is nominally the same process arrive at materially different results, without either being aware that anything differs.
A further contributor is that the gap is often actively concealed, not through dishonesty but through reasonable caution. Where a written procedure is treated as a compliance standard, an adaptation that improves the outcome while departing from the text is safer left unmentioned. The organisation has then created conditions in which its most useful operational knowledge is the knowledge least likely to be reported upward. Amy Edmondson’s research at Harvard Business School on psychological safety and learning behaviour in work teams, published in Administrative Science Quarterly, examined why teams differ in their willingness to raise what they know; her field study found that teams’ willingness to engage in learning behaviour tracked their sense of interpersonal safety rather than their confidence in their own effectiveness. Applied to this territory, it suggests that the visibility of operational reality is partly a function of what it is safe to say.
The costs are mostly deferred, which is why they attract so little attention until they arrive.
Undocumented practice is not transferable, so it leaves when its holder leaves. Organisations regularly discover the existence of a critical adaptation at the moment the person who performed it retires, resigns or takes extended leave, and performance in that area declines for reasons nobody can immediately account for. Undocumented practice also cannot be improved deliberately, because improvement requires knowing what is currently being done. And it cannot be extended to a new site or team, which is why replication of a successful operation so frequently produces something that resembles the original without performing like it.
There is a subtler cost. When the record of how work is done is unreliable, every subsequent analysis built on that record inherits the unreliability. Capacity assumptions, cost models, automation plans and improvement projects all take the documented process as their starting point. An organisation can conduct a rigorous exercise on a description that does not correspond to reality and arrive, rigorously, at the wrong place.
A common misconception deserves flagging here: that a large gap between documented and actual practice indicates indiscipline. Frequently it indicates the opposite — a workforce compensating competently for a description that no longer fits the conditions. The presence of adaptation is information about the environment, not only about the people in it. Related observations appear in why standard operating procedures stop being followed and the role of documentation in operational consistency.
Senior visibility of operations declines as organisations grow, and it declines faster than most leadership teams expect. This is not primarily a reporting failure. It is a structural consequence of distance, and of the substitutions organisations make to compensate for it.
In a small organisation, senior people see the work. They are physically near it, they encounter its exceptions directly, and their sense of how things are going is built from observation. As the organisation grows, that direct channel narrows and then closes. What replaces it is a representation of the work: reports, dashboards, metrics, exception summaries and periodic reviews. The representation is necessary — there is no alternative at scale — but it is a compression, and compressions lose things selectively rather than randomly.
What survives compression is what is countable, what is already anticipated, and what someone has an incentive to report. What does not survive is the unanticipated, the qualitative, the not-yet-nameable, and the awkward. This is why leadership teams are so often accurate about volumes and inaccurate about conditions.
The moment an organisation begins to manage through indicators, the indicators start to shape the behaviour they were built to observe. V. F. Ridgway set this out in Administrative Science Quarterly in 1956, in a paper examining single, composite and multiple performance measures. His documented examples remain striking: public employment interviewers assessed on the number of interviews they conducted worked through them quickly and placed few applicants; law-enforcement investigators working to a quota of eight cases a month selected easy cases as the month closed. Ridgway’s conclusion was not that measurement should be abandoned but that undue confidence in measures, without knowledge of their full effects, can leave the cure worse than the disease.
The relevance to operations is direct. An operational indicator is a proxy — handling time standing in for efficiency, defect counts standing in for quality, utilisation standing in for productivity. Proxies are useful precisely because they are simpler than the thing they represent, and that simplification is also their weakness. When the proxy becomes the object of management attention, effort migrates toward improving the proxy, which is not always the same as improving the outcome. Steven Kerr, writing in the Academy of Management Journal in 1975, described the general form of this: people work out what is actually rewarded and orient toward it, frequently at the expense of activities that are not rewarded.
None of this requires anyone to behave badly. It requires only that people respond rationally to what the organisation appears to value, which is what the measurement system communicates far more forcefully than any statement of priorities.
Operational information also degrades in transmission for reasons that have nothing to do with measurement design. Michael Beer and Russell Eisenstat, writing in what is now MIT Sloan Management Review, examined a set of mutually reinforcing barriers to strategy implementation and organisational learning — among them poor vertical communication and unclear or conflicting priorities — and observed that these barriers are frequently invisible to the senior team they most affect.
The mechanism is cumulative rather than dramatic. Each level in a reporting chain has a legitimate interest in presenting a resolved account rather than an unresolved one, and each summarises upward. Problems that are being worked on are reported as being worked on. Problems that are ambiguous are held pending clarity. Problems that would be embarrassing to escalate are addressed locally where possible. At each step, a small and defensible amount of texture is removed. By the third or fourth iteration, a situation that is genuinely uncertain on the ground has become, in the version that reaches the top, a settled matter with a modest variance.
This produces a specific and recognisable failure: a leadership team that is genuinely surprised by an operational problem that a great many people had known about for months. The surprise is usually read as a reporting failure and treated with more reporting — which frequently makes matters worse, because it increases the volume of representation without improving its fidelity. The territory is developed further in why operational visibility decreases as teams expand and how operational complexity hides inside everyday work.
There is a pattern in operations so consistent that it deserves to be treated as a structural feature rather than a local failing: improvement work is displaced by urgent work, and the displacement compounds.
Nelson Repenning and John Sterman of MIT Sloan examined this using system dynamics modelling alongside more than a dozen case studies, in a paper published in the California Management Review in 2001 under a title that captures the mechanism exactly — nobody ever gets credit for fixing problems that never happened. Their account describes what they termed a capability trap. Under pressure, managers default to working harder rather than working smarter, because effort produces visible results quickly while capability improvement produces results slowly and invisibly. The improvement that would reduce future pressure is therefore deferred in favour of the response that relieves present pressure, which increases future pressure, which makes the next deferral more likely.
The asymmetry in credit is central. A person who resolves a serious incident under time pressure is visible, valuable and frequently recognised. A person whose work meant the incident never occurred is, by definition, associated with nothing having happened. Organisations tend to reward the first and overlook the second, not by policy but by the structure of what is observable. Over time, this shapes where capable people direct their attention.
The trap has recognisable surface features, though it is rarely identified as a single phenomenon.
Improvement initiatives are launched with genuine commitment and then quietly fade — not cancelled, not evaluated, simply attended to less each month until they stop being mentioned. The people assigned to them are the same people the organisation calls on when something breaks, and something always breaks. Retrospective reviews are conducted after significant failures and produce sensible conclusions that are partially implemented. Operational teams describe themselves as permanently behind, and the description is accurate.
A related presentation is the reappearing problem. An issue is resolved, closes, and returns eight months later in a form similar enough to be recognisable and different enough to be treated as new. Where this recurs, it usually indicates that the earlier resolution addressed the presenting instance rather than the conditions producing it — a distinction that is easy to state and genuinely hard to act on when the presenting instance is causing damage today.
The intuitive response is capacity: the team is overloaded, so expand it. This sometimes works and frequently disappoints, for reasons connected to the first section of this page. Added people arrive into an existing structure of interfaces and handovers, and they bring their own coordination requirements with them. The additional capacity is real but partially consumed by the additional coordination it creates. Organisations that have added significant operational headcount without a corresponding improvement in throughput have generally encountered this, and the experience is disorienting because the arithmetic seemed obvious.
Capacity constraints are also frequently misidentified. A team that appears short of people may be absorbing rework generated elsewhere, or holding coordination load that belongs to an interface, or compensating for a documented process that no longer describes the work. Adding capacity in that situation raises the ceiling on absorbing the problem without altering the source. This is one of the reasons operational strain has a tendency to reappear in a different team after being relieved in the first.
Discussion of this territory frequently arrives at a familiar claim: that some large proportion of change and improvement initiatives — most often quoted as seventy per cent — fail. Mark Hughes of the University of Brighton examined this directly in the Journal of Change Management in 2011, tracing five separately published instances of the claim and finding no valid or reliable empirical evidence supporting any of them.
The figure is worth setting aside not because improvement work is easy, but because a number that cannot be traced to a methodology tells a leadership team nothing about its own situation. It functions instead as a licence for fatalism — if most attempts fail, the failure of any particular attempt requires no examination. The observable patterns described above are more useful than the statistic precisely because they are specific enough to be checked against a real organisation.
The research relevant to this territory comes from several traditions that developed largely independently — organisation design, performance measurement, system dynamics, safety science and organisational learning — and it is the convergence between them that gives the picture its weight.
The strongest signal concerns where the difficulty is located. Research by Donald Sull of MIT Sloan, Rebecca Homkes of London Business School and the LSE Centre for Economic Performance, and Charles Sull, published in Harvard Business Review in 2015 and based on a multi-year study of more than 250 companies and several thousand managers, found that only nine per cent of managers said they could rely on colleagues in other functions and units all of the time, with around half saying they could rely on them most of the time. Their broader conclusion was that the central execution problem is lateral rather than vertical: alignment processes in most firms function reasonably well, while coordination across units does not. For an operational context, that finding relocates the problem. It sits between well-run parts, not inside badly-run ones.
Galbraith’s information-processing account, published in Interfaces in 1974, supplies a reason why. If the volume of information an organisation must process during the work is driven by the uncertainty in its tasks, then growth — which reliably introduces variants, exceptions and new parties — raises the processing requirement regardless of how competently any individual unit operates. The load is a property of the configuration.
The measurement literature explains why this often goes unregistered. Ridgway’s 1956 examination of performance measures in Administrative Science Quarterly documented all three of the measurement approaches he examined producing undesirable consequences for overall performance, and Kerr’s 1975 analysis in the Academy of Management Journal set out the general principle that behaviour follows what is rewarded rather than what is stated. Together they suggest that an organisation’s indicators are not neutral instruments; they participate in producing the reality they report.
Safety and resilience research adds a further layer. The distinction between work as imagined and work as done, developed by Hollnagel and set out with Wears and Braithwaite in their 2015 white paper, holds that successful outcomes generally depend on situational adjustment rather than procedural compliance. That claim runs directly counter to the assumption embedded in most operational documentation, and it means that an organisation’s formal record of its own operations should be treated as an approximation of uncertain accuracy rather than as a description.
Repenning and Sterman’s modelling and case work in the California Management Review explains the persistence of all of it. If prevention is systematically under-credited relative to correction, then the capability that would reduce operational difficulty is systematically underinvested in, and the state of chronic strain is self-sustaining rather than transitional.
Two qualifications are worth stating plainly. First, none of this research was conducted on your organisation; it establishes that mechanisms exist and recur, not that any particular mechanism is operating in a particular case. Second, much of it is drawn from large firms and from specific sectors — manufacturing, healthcare, public administration — and the extent to which findings transfer to smaller or differently-structured organisations is a legitimate open question rather than a settled one. Research of this kind is better used to sharpen a hypothesis than to confirm one.
Consider an organisation running two production sites on the same documented process. Same procedure, same equipment specification, same training material, same reporting. One site consistently produces more rework than the other, and has done so for long enough that the difference has become an accepted characteristic of that site rather than a question about it.
The explanation that presents itself first is capability. The weaker site has less experienced staff, or supervision that is less attentive, or a local culture that has drifted. It is a reasonable inference, and it is the one most organisations in this position would begin with — partly because it is plausible and partly because it suggests an obvious response.
Observation of the actual work complicates it. Both sites follow the documented process. Neither is cutting corners, and neither is failing to do what the procedure requires. The difference is that the stronger site performs an additional check, introduced by a supervisor several years earlier, which is not in any document and which nobody at the other site knows exists. It has been passed on locally, by demonstration, to everyone who joined that site since.
What the situation then reveals is not a gap between two sites but a gap between the organisation and its own record. The documented process is genuinely identical at both locations. The actual process is not, and has not been for years. The organisation’s account of how the work is done describes neither site accurately — it under-describes one and over-describes the other, and the comparison it enabled was therefore never a comparison between like and like.
This is a common shape. The visible difference between two units invites an explanation in terms of the units, when the more consequential fact is often what the description of the work is failing to capture in both.
How much of what makes something work is not in the description of it?
“The record of how work is done is not the work. Organisations discover the distance between the two at inconvenient moments.” — Zoom Riddle
The scenario above is an invented composite constructed to illustrate a pattern. It does not describe any actual organisation.
These are offered as questions rather than as a checklist. They are open, and several have no clean answer.
Reflections from real work on organisations growing beyond their founding design.
GROWTH EXPOSES DESIGN
Building commercial infrastructure for an industrial manufacturer preparing for expansion shows that growth does…
THE SYSTEM SHOULD OUTLIVE THE FOUNDER
Preparing a fast-growing property network for franchising shows that a business can outgrow its founder’s…
READINESS UNDER UNCERTAINTY
Working with a specialist data-recovery and cyber-incident response business shows what marketing looks like when…
More on operational complexity: how consistency erodes as organisations grow.
Operational complexity describes the coordination burden an organisation carries as a consequence of how its work is arranged, rather than the difficulty of any individual task. It grows with the number of interfaces between people, teams, sites and systems, and because those interfaces multiply faster than the elements they connect, complexity typically rises faster than headcount. It is not a synonym for size or for disorder: a large organisation can be operationally simple, and a small one operating across many partial handovers can be operationally complex.
Consistency erodes because local adaptation is continuous while documentation is periodic. Conditions change in each part of the organisation at different rates, people adjust their work sensibly in response, and those adjustments spread as far as informal contact reaches rather than uniformly. Over time, units running what is nominally the same process are running materially different versions of it. The erosion is usually gradual and rarely visible at any single point, which is why it tends to be identified only when outcomes diverge enough to demand explanation.
Not usually. Research in safety and resilience — notably the distinction between work as imagined and work as done developed by Erik Hollnagel and colleagues — suggests that successful outcomes generally depend on people adjusting to conditions rather than following procedures exactly. A gap is normal and often functional. What matters more is whether the organisation knows the gap exists, whether the adaptations are visible enough to be protected and transferred, and whether the environment makes it safe to describe them.
Two mechanisms account for most recurrence. The first is that the resolution addressed the presenting instance rather than the conditions that produced it, so a similar instance arises later in a slightly different form. The second is that bottlenecks frequently sit at interfaces between parts of the organisation rather than within them; relieving pressure at one interface tends to reveal the next constraint rather than remove constraint altogether. Recurrence is therefore often informative rather than simply frustrating.
Sometimes, and less often than expected. Additional people arrive into an existing structure of handovers and bring coordination requirements of their own, so part of the added capacity is consumed by the coordination it creates. Apparent capacity shortages are also frequently something else — rework generated elsewhere, load belonging to an interface, or compensation for a process description that no longer matches the work. Where that is the case, additional capacity raises the organisation’s tolerance for the problem without altering its source.
Information degrades as it is summarised upward, and each individual summarisation is defensible. Problems being worked on are reported as being worked on; ambiguous problems are held pending clarity; problems that can be handled locally are handled locally. A situation that is genuinely unresolved on the ground can become, after several iterations, a settled item with a modest variance. Research on organisational learning has repeatedly found that barriers of this kind are least visible to the senior teams they affect most.
The figure is very widely repeated and cannot be traced to a stated methodology. Mark Hughes of the University of Brighton examined five separately published instances of the claim in the Journal of Change Management in 2011 and found no valid or reliable empirical evidence supporting any of them. The claim is best set aside — not because improvement work is easy, but because an untraceable number tells a leadership team nothing about its own circumstances.
Frequently not. Operations is where the consequences of decisions made elsewhere become physically observable, so operational symptoms often originate in unclear priorities, unowned decisions or measurement that has drifted from its purpose. This is one reason operational corrections fail to hold: the correction is applied where the symptom surfaced rather than where it began.
Beer, M., & Eisenstat, R. A. (2000). “The Silent Killers of Strategy Implementation and Learning.” Sloan Management Review, 41(4), 29–40.
Edmondson, A. C. (1999). “Psychological Safety and Learning Behavior in Work Teams.” Administrative Science Quarterly, 44(2), 350–383.
Galbraith, J. R. (1974). “Organization Design: An Information Processing View.” Interfaces, 4(3), 28–36.
Greiner, L. E. (1972). “Evolution and Revolution as Organizations Grow.” Harvard Business Review, 50(4), 37–46. Republished as an HBR Classic, 76(3), May–June 1998, 55–68.
Hollnagel, E., Wears, R. L., & Braithwaite, J. (2015). From Safety-I to Safety-II: A White Paper. Published jointly by the University of Southern Denmark, the University of Florida and Macquarie University.
Hughes, M. (2011). “Do 70 Per Cent of All Organizational Change Initiatives Really Fail?” Journal of Change Management. University of Brighton.
Kerr, S. (1975). “On the Folly of Rewarding A, While Hoping for B.” Academy of Management Journal, 18(4), 769–783.
Repenning, N. P., & Sterman, J. D. (2001). “Nobody Ever Gets Credit for Fixing Problems that Never Happened: Creating and Sustaining Process Improvement.” California Management Review, 43(4), 64–88.
Ridgway, V. F. (1956). “Dysfunctional Consequences of Performance Measurements.” Administrative Science Quarterly, 1(2), 240–247.
Rogers, P., & Blenko, M. W. (2006). “Who Has the D? How Clear Decision Roles Enhance Organizational Performance.” Harvard Business Review, 84(1), 52–61.
Sull, D., Homkes, R., & Sull, C. (2015). “Why Strategy Execution Unravels — and What to Do About It.” Harvard Business Review, 93(3), 57–66.
Services can be delivered remotely, and we accept requests from organisations wherever they are located.
